hslab.tex

隐马尔可夫模型源代码

%/* ----------------------------------------------------------- */
%/*                                                             */
%/*                          ___                                */
%/*                       |_| | |_/   SPEECH                    */
%/*                       | | | | \   RECOGNITION               */
%/*                       =========   SOFTWARE                  */ 
%/*                                                             */
%/*                                                             */
%/* ----------------------------------------------------------- */
%/*         Copyright: Microsoft Corporation                    */
%/*          1995-2000 Redmond, Washington USA                  */
%/*                    http://www.microsoft.com                */
%/*                                                             */
%/*   Use of this software is governed by a License Agreement   */
%/*    ** See the file License for the Conditions of Use  **    */
%/*    **     This banner notice must not be removed      **    */
%/*                                                             */
%/* ----------------------------------------------------------- */
%
% HTKBook - Steve Young  30/10/95
%

\newpage
\mysect{HSLab}{HSLab}

\newcommand{\hbut}[1]{\texttt{#1}}
\newcommand{\hbbut}[1]{\fbox{\texttt{#1}}}

\mysubsect{Function}{HSLab-Function}

\index{hslab@\htool{HSLab}|(}
\htool{HSLab} is an interactive label editor for manipulating speech label files. 
An example of using \htool{HSLab} would be to load a sampled waveform file, 
determine the boundaries of the speech units of interest and assign labels 
to them. Alternatively, an existing label file can be loaded and edited by
changing current label boundaries, deleting and creating new labels.
\htool{HSLab} is the only tool in the HTK package which makes use of the 
graphics library \htool{HGraf}.

When started \htool{HSLab} displays a window which is split into two parts: a
display section and a control section (see Fig~\href{f:hslab}). 
The display section contains the plotted speech waveform with the 
associated labels. The control section consists of a palette of buttons 
which are used to invoke the various facilities available in the tool. 
The buttons are laid out into three different groups depending on the 
function they perform. Group one (top row) contains buttons related to 
basic input/output commands. 
Group two (middle row) implements the viewing and record/playback functions. 
The buttons in group three (bottom row) are used for labelling. To invoke a 
particular function, place the mouse pointer onto the corresponding button 
and click once. All commands which require further interaction with the user 
after invocation will display guiding text in the message area telling the 
user what he or she is expected to do next. For example, to delete a label, 
the user will click on \hbut{Delete}, the message ``Please select label to delete'' 
will appear in the message area and the user will be expected to click in that 
part of the display section corresponding to the label to be deleted (not on 
the label itself).

A {\em marked region} is a slice of the waveform currently visible in the 
window. A region is marked by clicking on \hbut{Mark} and specifying two 
boundaries by clicking in the display section. 
When marked, a region will be displayed in inverse colours. In the presence 
of a marked region the commands \hbut{Play}, \hbut{Label} and \hbut{Label as} 
will be applied to the specified region rather than to the whole of the 
waveform visible on the screen. 
Part of the waveform can also be made into a marked region with the commands 
\hbut{Zoom Out} and \hbut{Select}. \hbut{Zoom Out} will take the user back 
to the previous level of magnification and the waveform being displayed before 
the execution of the command will become a marked region. 
\hbut{Select} will make the part of the waveform corresponding to a particular 
label into a marked region. This can be useful for playing back existing labels. 

Labelling is carried out with \hbut{Label } and \hbut{Label as}. \hbut{Label}
will assign {\em The Current Label} to a specified slice of the waveform,
whilst \hbut{Label as} will prompt the user to type-in the labelling string. 
{\em The Current Label} is shown in the button in the bottom right corner of 
the control section.  It defaults to ``Speech'' and it can be changed by clicking 
on the button it resides in. 
Multiple alternative transcriptions are manipulated using the \hbut{Set [?]} and 
\hbut{New} buttons. The former is used to select the desired transcription, the 
latter is used to create a new alternative transcription.

\centrefig{hslab}{120}{HSLab display window}

\mysubsect{Use}{HSLab-Use}

\htool{HSLab} is invoked by typing the command line
\begin{verbatim}
   HSLab [options] dataFile
\end{verbatim}
where \texttt{dataFile} is a data file in any of the supported formats with a 
\texttt{WAVEFORM} sample kind.   If the given data file does not exist, then
\htool{HSLab} will assume that a new file is to be recorded with this name.

The available options for \htool{HSLab} are

\begin{optlist}
  \ttitem{-a} With this switch present, the numeric part of the global labelling
  string is automatically incremented after every \texttt{Label} operation.

  \ttitem{-i file} This option allows transcription files to be output to the
  named master label file (MLF). 

  \ttitem{-n} Normally \htool{HSLab} expects to load an existing label file
  whose name is derived from the speech data file. This option tells 
  \htool{HSLab} that a new empty transcription is to be created for the loaded 
  data-file.

  \ttitem{-s string} This option allows the user to set the string displayed
  in the ``command'' button used to trigger external commands.

\stdoptF
\stdoptG
\stdoptI
\stdoptL
\stdoptX
\end{optlist}
\stdopts{HSLab}

The following is a summary of the function of each \htool{HSLab}
button.

\begin{optlist}
  \ttitem{\hbbut{Load}} Load a speech data file and associated transcription.
      If changes have been made to the currently loaded transcription 
      since it was last saved the user will be prompted to save these 
      changes before loading the new file.

  \ttitem{\hbbut{Save}} Save changes made to the transcription into a specified
      label file.

  \ttitem{\hbbut{About}} Print information about HSLab.

  \ttitem{\hbbut{Quit}} Exit from \htool{HSLab}. If alterations have been made 
      to the currently loaded transcription since it was last saved, the user 
      will be prompted to save these changes before exiting.

  \ttitem{\hbbut{Command}} This button is used to trigger an external command which 
      can process the waveform file currently loaded in \htool{HSLab}. This 
      is accomplished by setting the environment variable \texttt{HSLABCMD} to the 
      shell command required to perform the processing. When the \texttt{Command}
      button is pressed, any occurrence of \texttt{\$} in the shell command is 
      replaced by the name of the currently loaded waveform file. Note that 
      only the filename without its extension is substituted. The string 
      displayed in the ``command'' button can be changed using the \texttt{-s} 
      option.

  \ttitem{\hbbut{Mark}} Mark a region of the displayed waveform. The user 
      is prompted to specify the start and the end point of a region with
      the mouse pointer. The marked region will be displayed in inverse 
      colours. Only one region can be marked at a time.

  \ttitem{\hbbut{Unmark}} Unmark a previously marked region.

  \ttitem{\hbbut{<--}} Scroll the display to the left.

  \ttitem{\hbbut{-->}} Scroll the display to the right.

  \ttitem{\hbbut{Z.In}} Zoom into a part of the displayed waveform. If there is 
      a currently marked region then that region will be zoomed into, 
      otherwise, the user will be prompted to select a slice of the 
      waveform by specifying two points using the mouse pointer.

  \ttitem{\hbbut{Z.Out}} Restore the previous viewing level.  

  \ttitem{\hbbut{Restore}} Display the complete waveform into the window. Any 
      marked regions will be unmarked.  

  \ttitem{\hbbut{Play}} If there is a marked region of the waveform then
      that portion of the signal will be played through the internal speaker.
      Otherwise, the command will apply to the waveform visible on the 
      screen.

  \ttitem{\hbbut{Rec}} This initiates recording from the audio input device. The maximum
      duration of a recording is limited to 2 mins at 16KHz sampling rate. Two bar-graphs 
      are displayed: the first (red) shows the number of samples recorded, the second
      bar (green) displays the energy of the incoming signal. Once pressed, the
      \texttt{Rec} button changes into \texttt{Stop} which, in turn, is used to terminate
      the operation. When finished, the audio data stored in the buffer is written
      out to disk. Each recording is stored in alternating files 
      \texttt{dataFile\_0} and \texttt{dataFile\_1}.

  \ttitem{\hbbut{Pause}} Clicking on this button pauses/un-pauses the recording operation.

  \ttitem{\hbbut{Volume}} This button is used to select the playback volume of the
      audio device.

  \ttitem{\hbbut{x1}} This button selects the current level of waveform magnification.
      The available factors are $\times 1$,  $\times 2$,  $\times 4$,  $\times 8$, 
       $\times 16$, and  $\times 32$.

  \ttitem{\hbbut{Label}} If a marked region exists, then the waveform contained in the 
      region will be labelled with {\em The Current Label}. Otherwise, the 
      command will be applied to the waveform visible on the screen.

  \ttitem{\hbbut{Labelas}} Same as above, however, the user is prompted to type in 
      the labelling string.

  \ttitem{\hbbut{Delete}} Delete a label.

  \ttitem{\hbbut{Edit}} Edit the string of a label.

  \ttitem{\hbbut{Select}} Select a label as a marked region.

  \ttitem{\hbbut{Adjust}} Adjust the boundaries of a label.  To select the label 
      boundary to adjust, click in the display near to the label boundary.

  \ttitem{\hbbut{Set [?]}} This button is used to select the current alternative 
      transcription displayed and used in \htool{HSLab}.

  \ttitem{\hbbut{New}} Creates a new alternative transcription. If an empty alternative 
      transcription already exists, then a new transcription is not created.

  \ttitem{\hbbut{Undo}} Single level undo operation for labelling commands.

  \ttitem{\hbbut{Speech}} Change the current labelling string (the button in the 
  bottom right of the control area).
\end{optlist}

The following ``mouse'' shortcuts are provided. To mark a region position the pointer at
one of the desired boundaries, then press the left mouse button and while holding it down
position the pointer at the other region boundary. Upon releasing the mouse button the 
marked region will be hilighted. To play a label position the mouse cursor anywhere within
the corresponding label ``slice'' in the label area of the display and click the left
mouse button.

\mysubsect{Tracing}{HSLab-Tracing}
\htool{HSLab} does not provide any trace options.
\index{hslab@\htool{HSLab}|)}


%%% Local Variables: 
%%% mode: latex
%%% TeX-master: "../htkbook"
%%% End: